International Comparisons of Poverty in South Asia

International Comparisons of Povertyin South Asia

TM Tonmoy Islam, David Newhouse, and Monica Yanez-Pagans∗

This paper explores the methodological differences underlying the constructionof the national consumption aggregates that are used to estimate internationalpoverty rates for South Asian countries. The analysis draws on a regional datasetof standardized consumption aggregates to assess the sensitivity of internationalpoverty rates to the items included in the national consumption aggregates.A key feature of the standardized aggregate is that it includes the reportedvalue of housing rent for urban Indian homeowners. Using the standardizedconsumption aggregates reduces the international poverty rate in South Asiaby 1.3 percentage points, impacting the status of about 18.5 million people.Comparing standardized and nonstandardized monetary welfare indicators toother nonmonetary indicators suggests that the latter are more consistentwith the standardized consumption aggregates. Overall, the results stronglysuggest that harmonizing the construction of welfare measures, particularlythe treatment of imputed rent, can meaningfully improve the accuracy ofinternational poverty comparisons.

Keywords: Bangladesh, consumption aggregate, imputed rent, India, povertymeasurement, South AsiaJEL codes: I32

I. Introduction

The World Bank has set an ambitious target to eradicate extreme povertyglobally by 2030. Evaluating progress toward this goal requires a solid datainfrastructure. In April 2013, the World Bank set two goals to guide its work inthe coming years—the first is to eradicate extreme poverty from the world by 2030,while the second is to promote shared prosperity. The first goal will be achieved ifthe incidence of extreme poverty falls below 3% by 2030, while the second goal

∗TM Tonmoy Islam (corresponding author): Elon University, North Carolina, United States.E-mail: [email protected]; David Newhouse: World Bank, Washington, DC, United States. E-mail:[email protected]; Monica Yanez-Pagans: World Bank, Washington, DC, United States. E-mail:[email protected]. We thank Nicola Amendola and Giovanni Vecchi for their support in producinginputs for the preparation of this paper, Julian Diaz-Gutierrez and Yurani Arias-Granada for their excellent researchassistance, and Benu Bidani and Martin Rama for their financial support through the South Asia Regional Datafor Goals Program. We also thank the managing editor and the anonymous referees for helpful comments andsuggestions. The Asian Development Bank (ADB) recognizes “China” as the People’s Republic of China and“Vietnam” as Viet Nam. The usual ADB disclaimer applies.

Asian Development Review, vol. 38, no. 1, pp. 142–175https://doi.org/10.1162/adev_a_00161

© 2021 Asian Development Bank andAsian Development Bank Institute.

Published under a Creative CommonsAttribution 3.0 International (CC BY 3.0) license.

International Comparisons of Poverty in South Asia 143

can be attained by increasing the average welfare of those who earn at the 40thpercentile or below in each country (Jolliffe et al. 2014). To achieve the first goal, itwill be important to allocate resources to countries where extreme poverty is mostprevalent. However, poverty comparisons across countries are partly influenced bythe quirks of each country’s surveys. A thorough analysis is needed to understandthe similarities and differences of these survey data collection procedures and howmuch they affect reported poverty rates.

Extreme poverty is measured using the international poverty line, currentlyset at $1.9 per person per day in 2011 dollars (hereafter: international povertyrate).1 Anyone consuming less than that amount, or earning less than that amountin countries that use income rather than consumption as their primary welfaremeasure, is assumed to be extremely poor. In addition, each country sets its ownofficial poverty line to assess national poverty, with the line depending on the levelsof consumption in each country. Therefore, official poverty lines and the subsequentnational poverty rates generated using these lines are not comparable acrosscountries (hereafter: national poverty rates). To generate international povertyestimates, the World Bank has created an international poverty line, which is appliedconsistently to all countries to monitor extreme poverty. A poverty line of $1per day was first estimated in 1990 and has been repeatedly updated to take intoaccount changes in the purchasing power parity of countries (Ravallion, Chen, andSangraula 2009).2 The last update of the international poverty line was done in2015, when it was raised to $1.9 per person per day according to the 2011 values ofpurchasing power parity exchange rates (Ferreira et al. 2016).

Globally, the extreme poor are concentrated in sub-Saharan Africa and SouthAsia. Table 1 shows the number and proportion of the extreme poor in 2013 byglobal region. On average, 12.6% of the world’s population, or about one in eightpeople, lives in extreme poverty. Sub-Saharan Africa and South Asia have thehighest and second-highest number and proportion of the world’s extreme poor,respectively, with 50.7% and 33.4% of the world’s extreme poor living in these tworegions.

A recent report by the Commission on Global Poverty to improve globalpoverty monitoring highlights the considerable uncertainty in global povertyestimates. Since adopting the twin goals mentioned above, the World Bank hasdevoted considerable attention to improving its measure of extreme poverty. As

1This line was calculated in the following way: (i) the national poverty lines of 15 poor countries wereselected; (ii) those poverty lines were inflated to 2011 levels using the consumer price index (CPI) of those countries;(iii) the inflated poverty lines were then converted to United States (US) dollars using 2011 purchasing power parity(PPP); and (iv) those US dollar-denominated poverty lines at PPP were averaged to come up with a new poverty line,which was close to $1.9 per person per day. The countries in the reference category are Chad, Ethiopia, the Gambia,Ghana, Guinea-Bissau, Malawi, Mali, Mozambique, Nepal, Niger, Rwanda, Sierra Leone, Tajikistan, Tanzania, andUganda (Ferreira et al. 2016).

2For more information, consult the World Bank. 1990. World Development Report 1990. https://openknowledge.worldbank.org/bitstream/handle/10986/5973/WDR%201990%20-%20English.pdf?sequence=5.

https://openknowledge.worldbank.org/bitstream/handle/10986/5973/WDR%201990%20-%20English.pdf?sequence=5

144 Asian Development Review

Table 1. International Poverty Rate at $1.9 per Day andNumber of Extreme Poor by Region, 2013

Region

InternationalPoverty Rate

(%)

Number ofExtreme Poor

(million)

East Asia and Pacific 3.54 71.02Europe and Central Asia 2.15 10.30Latin America and the Caribbean 5.40 33.59Middle East and North Africa na naSouth Asia 15.09 256.24Sub-Saharan Africa 40.99 388.72World average 12.55 766.01

na = not available.Note: All amounts are based on 2011 purchasing power parity.Source: Authors’ estimates based on World Bank. PovcalNet. http://iresearch.worldbank.org/PovcalNet/povOnDemand.aspx (accessed September 15, 2017).

part of this effort, the World Bank established the Commission on Global Poverty,chaired by Sir Anthony Atkinson, to come up with different recommendations onhow to improve the measurement and monitoring of global extreme poverty. Thecommission’s report highlighted shortcomings in the global poverty measurementinfrastructure in detail and offered several suggestions on how to improve themonitoring of global extreme poverty.

An important recommendation provided by the Commission on GlobalPoverty’s report is to calculate and include the “total error” of the poverty estimatesfor each country (World Bank 2016 and 2017). Specifically, Recommendation 5suggests that poverty estimates should be based on a total error approach, evaluatingthe possible sources and magnitude of error, particularly nonsampling error andthe error introduced by the process of determining the international poverty line.There are many factors that can affect the total error of international povertyrate estimates—such as incomplete coverage of the country’s population, errorsin measuring consumption data, errors in calculating the poverty line, use of theconsumer price index (CPI) to deflate prices in a manner that may not be consistentwith the consumption patterns of the poor, and geographic differences in prices.Because the total error may be significant, the report recommended that the WorldBank provide a margin of error to help policy makers understand the accuracy ofthe reported extreme poverty numbers. Although it would be difficult to implement,the World Bank concluded that reporting total error with poverty estimates is oneof the report’s most important recommendations.

South Asia is a useful laboratory to study how methodological differencesin poverty measurement can contribute to total error. This is because South Asiais home to about one-third of the world’s extreme poor, but the region consists ofonly eight countries, making the analysis both significant from a global perspectiveand tractable. In South Asia, all eight countries compare consumption per capita

http://iresearch.worldbank.org/PovcalNet/povOnDemand.aspx


against the poverty line to identify the international extreme poverty status of anindividual.3

This paper provides new evidence from South Asia on how differences inthe construction of the welfare measure in each country contributes to total error ininternational poverty measurement. To study this in detail, we look at how countriesin the region construct the consumption aggregate to assess poverty. A key source oftotal error is the collection and aggregation of household-level data from differentcountries. While all countries collect household data to measure poverty, there aredifferences in the methodology and the content included in the survey questionnaire,which contributes to the total error of the subsequent international poverty rateobtained for each country. For example, the list of goods included in the surveysis not consistent. Additionally, the steps in the methodology used to construct theconsumption aggregate vary from country to country, which then adds to the totalerror of the point estimates of the international extreme poverty rate.

We examine the following three aspects of the construction of consumptionaggregates to see how each of them contributes to total error: (i) samplingand survey design, (ii) spatial deflation and intertemporal deflation, and(iii) construction of the nominal consumption aggregate. The core of the papercompares poverty rates using the national consumption aggregates, which currentlyform the basis for poverty measurement, to standardized consumption aggregatesthat attempt to adjust for differences in the three aspects described above.This exercise provides an assessment of the extent to which the internationalpoverty rates depend on methodological differences in the construction of welfareaggregates across countries.

The remainder of this paper is organized as follows. Section II lists thedata sources and international poverty rates for the eight South Asian countries.Section III explains the differences in sampling and design across the surveys usedto estimate poverty in these countries. Section IV analyzes how spatial deflationaccounts for cost-of-living differences and how intertemporal deflation affectsinternational poverty estimates. Section V assesses the differences in estimating anational consumption aggregate across countries. Section VI examines the extentto which the international poverty rate in each country is correlated with othernonmonetary dimensions of well-being. Section VII concludes the paper.

II. Data and Poverty Rates

National and international poverty rates are estimated using nationallyrepresentative household-level surveys that collect detailed food and nonfoodconsumption data. Table 2 shows the household surveys that are used by the

3This contrasts with what most countries do in Latin America and the Caribbean and in Eastern Europe andCentral Asia, which use per-income-based measures to monitor poverty rather than consumption-based ones.


Table 2. South Asian Household-Level Surveys Used to Estimate Poverty

Country Survey Year

Afghanistan Living Conditions Survey (ALCS) 2012Bangladesha Household Income and Expenditure Survey (HIES) 2010Bhutan Living Standards Survey (LSS) 2012India National Sample Survey 68th Round (NSS) 2011Maldivesb Household Income and Expenditure Survey (HIES) 2009/2010Nepal Living Standards Survey (LSS) 2010Pakistan Pakistan Social and Living Standards Measurement (PSLM) 2011Sri Lanka Household Income and Expenditure Survey (HIES) 2012

aBangladesh conducted the latest round of the HIES in 2016–2017.bMaldives conducted the latest round of the HIES in 2016, but the data were not yet available as ofOctober 2018.Source: South Asia Harmonized Micro Dataset (accessed September 15, 2017).

Table 3. International Extreme Poverty Headcount for South Asian Countries

Country Year

InternationalPoverty Rate

(%)

Number ofExtreme Poor

(million)

GNI perCapita

(2011 PPP)

Afghanistan 2011 na $1,731.70Bangladesh 2010 18.51 27.49 $2,783.56

Rural 2010 22.70 24.89Urban 2010 6.70 2.60

Bhutan 2012 2.17 0.01 $6,452.00India 2011–2012 21.56 239.10 $4,594.20

Rural 2011–2012 24.84 196.70Urban 2011–2012 13.38 42.39

Maldives 2009 7.25 0.02 $9,714.40Nepal 2010 14.89 4.20 $2,053.40Pakistan 2011–2012 7.93 10.26 $4,516.50Sri Lanka 2012–2013 1.92 0.38 $9,121.40

GNI = gross national income, na = not available, PPP = purchasing power parity.Sources: Authors’ estimates based on South Asia Harmonized Micro Dataset (accessed September15, 2017); GNI per capita obtained from World Bank. World Development Indicators. https://datacatalog.worldbank.org/dataset/world-development-indicators (accessed September 15, 2017).

eight countries in South Asia to estimate poverty rates. Living standards varyconsiderably among countries in South Asia. Table 3 lists the international extremepoverty rates of each South Asian country, along with their gross nationalincome (GNI) per capita and number of extreme poor. The international extremepoverty rate of each country has been determined by calculating the proportion ofindividuals whose per capita consumption aggregate is lower than the internationalpoverty line. The GNI per capita for all South Asian countries is below $10,000,ranging from about $1,700 to about $9,700. Maldives and Sri Lanka have thehighest and second-highest GNI per capita, respectively, while Afghanistan andNepal have the lowest and second-lowest, respectively. International extreme

https://datacatalog.worldbank.org/dataset/world-development-indicators


poverty rates also vary considerably in the region, largely in line with the patternsobserved for GNI per capita, except for India. In Sri Lanka and Bhutan, theinternational extreme poverty rate is less than 3%. Pakistan and Maldives have aninternational extreme poverty rate in the range of 7%–8%. Nepal and Bangladeshhave the highest international extreme poverty rates, ranging from 14% to 19%.India is one of the few countries in the world for which the World Bank traditionallyreports international poverty rates by urban and rural regions; in rural and urbanareas of India, 24.8% and 13.4% of the population live below the internationalextreme poverty line, respectively.4

According to official figures, the international poverty rate is the highest inIndia among the eight countries in the region. India also has the largest numberof extreme poor among these countries, 239 million, which is about 28% of theworld’s extreme poor.5 By far, most of the extreme poor in India are from ruralareas. Bangladesh has the second-highest number of international extreme poor inSouth Asia (27 million) and Pakistan the third highest (10 million).

India’s high rate of poverty may partly reflect methodological differences inthe way in which countries construct their national consumption aggregates. Thehigher poverty rate in India compared to Bangladesh, for instance, is inconsistentwith the GNI per capita metric, which is almost $2,000 higher for India thanBangladesh. This suggests that differences in the way in which consumptionaggregates are constructed in each country might contribute to the total error. Tostudy how the national consumption aggregates are constructed, we look at the latestrounds of household surveys available for these countries.

III. Assessing Differences in Sampling and Survey Design

In this paper, we examine eight aspects of sampling and survey design thatare directly related to poverty measurement: (i) sampling design, (ii) monetarywelfare measure, (iii) food consumption questionnaire and data collection methods,(iv) self-production and meals outside home, (v) nonfood durables, (vi) durables,(vii) housing expenditures, and (viii) health and education expenditures.

Sampling Design

Table 4 presents a summary of the sampling design for each of theeight countries in South Asia. With few exceptions, household-level surveysused to measure poverty in the region are nationally representative. Afghanistan,Bangladesh, and India do not survey all regions within the borders of their

4The other countries for which international poverty rates are reported separately for urban and rural areasare the People’s Republic of China (PRC) and Indonesia.

5This proportion is calculated using PovcalNet data (accessed September 15, 2017).

148 Asian Development ReviewTa

ble

4.Su

mm

ary

ofSa

mpl

ing

Des

igns

inH

ouse

hold

-Lev

elSu

rvey

sU

sed

toM

easu

reP

over

ty

Afg

hani

stan

Ban

glad

esh

Bhu

tan

Indi

aM

aldi

ves

Nep

alP

akis

tan

SriL

anka

Nat

iona

llyre

pres

enta

tive?

Yes

aY

esb

Yes

cN

odY

ese

Yes

fY

esg

Yes

Sam

plin

gfr

ame

Pre

cens

usho

useh

old

list

ing,

cond

ucte

dbe

twee

n20

03an

d20

05

Popu

lati

onan

dH

ousi

ngC

ensu

s(2

001)

Popu

lati

onan

dH

ousi

ngC

ensu

s(2

005)

Rur

al:I

ndia

nC

ensu

sV

illa

ges

(200

1);U

rban

:U

rban

Fram

eS

urve

y

Popu

lati

onan

dH

ousi

ngC

ensu

s(2

006)

Nat

iona

lPo

pula

tion

Cen

sus

(200

0)

Urb

an:F

BS

’ur

ban

fram

e(2

003)

.Rur

al:

Popu

lati

onC

ensu

s(1

998)

Cen

sus

ofPo

pula

tion

and

Hou

sing

(201

1)

Res

pons

era

te89

%na

Urb

an:9

2%R

ural

:97%

cna

90%

e97

%f

na81

%h

Rur

alho

useh

old

sam

ple

size

86.5

%59

.0%

48.5

%59

.0%

64.3

%i

79.1

%60

.5%

77.4

%j

20,8

2812

,240

8,96

810

1,66

21,

832

7,18

016

,341

20,5

40

Sam

plin

gra

tio

0.64

0.04

7.01

0.04

3.71

0.12

0.08

0.40

Indi

vidu

alsa

mpl

esi

ze15

9,22

455

,580

39,8

2546

4,96

011

,588

34,8

1510

8,93

380

,534

Ref

eren

cepe

riod

(sur

vey)

Apr

2011

–A

ug20

12(1

)Fe

b20

10–

Jan

2011

b20

12c

Jul2

011–

Jun

2012

dS

ep20

09–

Sep

2010

e20

10f

Jul2

010–

Jun

2011

gJu

l201

2–Ju

n20

13h

FB

S=

Fede

ralB

urea

uof

Sta

tist

ics,

na=

nota

vail

able

.a N

atio

nal

Ris

kan

dV

ulne

rabi

lity

Ass

essm

ent

2011

–201

2,A

fgha

nist

anL

ivin

gC

ondi

tion

sS

urve

yR

epor

t.H

elm

and

and

Kho

stpr

ovin

ces

wer

eex

clud

edfr

omco

nsum

ptio

nin

form

atio

n.bS

tati

stic

alD

ivis

ion

ofth

eM

inis

try

ofP

lann

ing,

2011

Ban

glad

esh

Hou

seho

ldIn

com

ean

dE

xpen

ditu

res

Sur

vey:

Key

Find

ings

and

Rep

ort.

c Asi

anD

evel

opm

entB

ank,

2013

Bhu

tan

Liv

ing

Sta

ndar

dS

urve

y20

12R

epor

t.dA

llS

tate

san

dU

nion

Terr

itor

ies

exce

ptA

ndam

anan

dN

icob

arIs

land

san

dth

ere

mot

ear

eas

ofN

agal

and.

Indi

anC

entr

alS

tati

stic

alO

ffice

:Not

eon

Sam

ple

Des

ign

and

Est

imat

ion

Pro

cedu

reof

the

Nat

iona

lSam

ple

Sur

vey

68th

Rou

nd.

e Mal

è:S

ep.2

009–

Jan.

2010

;Ato

lls:

Feb.

2010

–Sep

.201

0.D

epar

tmen

tof

Nat

iona

lPla

nnin

g20

12,H

ouse

hold

Inco

me

and

Exp

endi

ture

sS

urve

y20

09–2

010.

f Cen

tral

Bur

eau

ofS

tati

stic

s,20

11N

epal

Liv

ing

Sta

ndar

dS

urve

y:H

ighl

ight

s.gS

tati

stic

Div

isio

nof

the

Gov

ernm

ento

fPa

kist

an,P

akis

tan

Soc

iala

ndL

ivin

gS

tand

ards

Mea

sure

men

tRep

ort.

hS

riL

anka

Dep

artm

ento

fC

ensu

san

dS

tati

stic

s,20

15H

ouse

hold

Inco

me

and

Exp

endi

ture

sS

urve

y.i A

toll

sar

eco

nsid

ered

rura

lare

as.

j The

reis

anad

diti

onal

5%in

the

esta

tese

ctor

,whi

chco

nsis

tsof

tea

plan

tati

ons.

Sou

rce:

Aut

hors

’es

tim

ates

base

don

Sou

thA

sia

Har

mon

ized

Mic

roD

atas

et(a

cces

sed

Sep

tem

ber

15,2

017)

.


respective countries. In 2011–2012, Afghanistan excluded the provinces ofHelmand and Khost from the survey for poverty measurement.6 These twoprovinces had an estimated population of 864,600 (Helmand) and 537,800 (Khost)in 2012 (Government of Afghanistan, Central Statistics Organization 2017). Thetotal population of Afghanistan in 2012 was about 24.8 million, so these twoprovinces combined represented around 5.7% of the population. Bangladesh didnot traditionally include the slum population as part of the sampling frame for theHousehold Income and Expenditure Survey (HIES) until 2016–2017. Accordingto the 2012 Bangladesh Bureau of Statistics’ Census of Slum Areas and FloatingPopulation, the slum population was about 2.22 million, which corresponds to5.5% of the total population in urban areas (Government of Bangladesh, Bureau ofStatistics 2012).7 The 68th round of the National Sample Survey (NSS) from Indiaexcluded from its sampling frame the remote areas of Nagaland and Andaman andNicobar Islands. The population of Andaman and Nicobar Islands is 0.38 million,while that of Nagaland is 1.98 million. Given that India had a population of 1,247million in 2011, the share of the total population excluded from the survey was just0.2%.

Sample sizes vary widely across household-level surveys used to measurepoverty. Maldives surveys about 1,800 households, while India surveys over100,000. The range of individuals covered by these household surveys varies from11,500 in Maldives to over 464,000 in India. This translates to a wide span ofsampling ratios—from 0.04% in Bangladesh to 7% in Bhutan.

Monetary Welfare Measure

Countries in the region use broadly similar methods to measure per capitaconsumption aggregates. All countries in South Asia use consumption rather thanincome to measure poverty. To estimate international poverty, total householdconsumption is divided by the number of individuals in the household to get a percapita estimate. This matches the methodology used by all countries in the region toestimate national per capita consumption aggregates, except for Pakistan. Pakistanuses per capita equivalence scales when measuring national poverty rates, thoughits international poverty rate is calculated using a simple per capita consumptionaggregate metric.

6The provinces of Helmand and Khost were included in the household survey, but these are not used toestimate poverty as there are issues with consumption data quality in these two provinces.

7There is a longstanding debate in Bangladesh about the size of the slum population. The Census andMapping of Slums collected by the Center for Urban Studies in 2009 estimated that 35.2% of the urban population inBangladesh lived in slums, while the UN-Habitat Global Report on Human Settlements released in 2013 estimatedthe proportion of urban slum dwellers in 2009 at 61.6%.


Food Consumption Questionnaire and Data Collection Methods

The questionnaire design methodologies for recording the data varyconsiderably across countries in the region. A comprehensive study analyzingsurvey questionnaires from 100 low- and middle-income countries noted the largevariation in the way surveys are designed in each country (Smith, Dupriez, andTroubat 2014). This variation is also apparent across the South Asian householdsurvey questionnaires. Table 5 lists the contents of the questionnaires from SouthAsian countries, which show significant differences in the way food consumptiondata are collected.

An important source of differences is the number of food items in theconsumption questionnaire. Surveys with more food items listed tend to elicithigher levels of consumption, which lowers the reported poverty rate (Lanjouwand Lanjouw 2001). Pakistan has the lowest number of food items listed in thesurvey at 69, while Sri Lanka has the highest at 227. Afghanistan, Maldives, Nepal,and Pakistan all have fewer than 100 food items listed in their survey, while India,Bangladesh, and Bhutan have 143, 141, and 130 items listed, respectively. Similarly,there is a large variation in the number of nonfood items listed in these surveys.Among nonfood expenditures, Nepal asks its respondents to recall only 95 items,Pakistan and Sri Lanka ask for 99 and 97 items, respectively, while Maldives lists483 nonfood items for households to recall. India, Bangladesh, and Bhutan asktheir households to recall whether they consumed 338, 221, and 122 nonfood items,respectively.

The collection of consumption data also varies from country to country.Consumption data are generally collected using two methods: (i) diary, wherethe household records all consumption data over a certain period in a notebook;or (ii) recall, where the household lists what they consumed over the past fewdays from memory. While the diary may seem to be a better method, in practice,interviewers often assist in completing the diary, effectively blurring the linebetween the two methods (Beegle et al. 2012). Sri Lanka uses both the diary andrecall methods to collect consumption data, Maldives uses diary for all items, whileother countries use recall to collect the data.

The length of consumption recall is another source of differences.Respondents are asked to recall their food consumption for the day, last week, last 2weeks, and last month. For food items, Bangladeshi enumerators spend 2 weeks inthe primary sampling unit, visiting each household seven times in a 2-week period.During each visit, the household is asked about their food consumption in the past 2days, so this covers a total of 14 days of food consumption. In addition, informationabout spice consumption in Bangladesh is collected once a week during those 2weeks. Bhutan asks the respondents to recall the last week, last month, and last yearof food consumption. India and Maldives ask the households to recall food itemsconsumed in the last month. Nepal asks for 1 week to 1 month of recall depending


Tabl

e5.

Sum

mar

yof

the

Con

sum

ptio

nQ

uest

ionn

aire

Des

ign

Afg

hani

stan

Ban

glad

esh

Bhu

tan

Indi

aM

aldi

ves

Nep

alP

akis

tan

SriL

anka

Food

expe

ndit

ures

(No.

ofit

ems)

9114

113

014

392

7469

227a

Dia

ryve

rsus

reca

llR

ecal

lR

ecal

lR

ecal

lR

ecal

lD

iary

Rec

all

Rec

all

Dia

ryan

dre

call

Ref

eren

cepe

riod

(Foo

dco

nsum

ptio

n)D

aily

and

7da

ysD

aily

and

wee

kly

(for

2w

eeks

)

7da

ys,1

mon

th,a

nd1

year

1m

onth

1m

onth

1w

eek

per

typi

calm

onth

2w

eeks

per

mon

th1

wee

k

Food

quan

titi

esav

aila

ble

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

b

Non

food

expe

ndit

ures

(No.

ofit

ems)

3822

112

233

848

395

9997

c

Ref

eren

cepe

riod

(Non

food

expe

ndit

ures

)1

and

12m

onth

s1,

3,an

d12

mon

ths

1an

d12

mon

ths

1an

d12

mon

ths

1an

d12

mon

ths

1an

d12

mon

ths

1an

d12

mon

ths

1,6,

and

12m

onth

s

CO

ICO

P=

clas

sifi

cati

onof

indi

vidu

alco

nsum

ptio

nac

cord

ing

topu

rpos

e.a T

his

incl

udes

six

liqu

oran

dto

bacc

oC

OIC

OP

code

sam

ong

food

expe

ndit

ures

.bU

nits

ofm

easu

rear

esp

ecifi

edin

the

ques

tion

nair

ebu

tnot

inth

eda

tase

ts.

c Toba

cco

and

drug

sex

clud

ed.

Sou

rce:

Aut

hors

’es

tim

ates

base

don

Sou

thA

sia

Har

mon

ized

Mic

roD

atas

et(a

cces

sed

Sep

tem

ber

15,2

017)

.


on the item, and Pakistan asks for between 2 weeks and 1 month of recall. Lastly,in Sri Lanka, enumerators visit the household three times during a week and askfor information about food consumption through a diary, while the information onnonfood items is collected by recall. When it comes to nonfood items, the length ofrecall is 1 month and 12 months for Afghanistan, Bhutan, India, Maldives, Nepal,and Pakistan. Bangladesh asks respondents to recall the last 1, 3, and 12 months ofnonfood consumption, while Sri Lanka asks for the last 1, 6, and 12 months.

Recall length has a large effect on the magnitude of the national consumptionaggregates. For example, when India changed the length of recall of consumptiongoods from 30 days to 7 days, consumption numbers reported by households wentup, and poverty rates fell by half (Deaton 2005). This simple change in the methodof collecting data “lifted” 175 million Indians out of poverty. Similarly, Beegle et al.(2012) found that changing the recall period from 1 week to 2 weeks in Tanzania,while holding other things equal, increased the poverty rate from 55% to 63%.Joliffe (2001) and Gibson, Huang, and Rozelle (2001) also show that poverty andinequality measures are significantly sensitive to the income recall period.8

Self-Production and Meals Outside Home

Another difference in the construction of the national consumptionaggregates is the treatment of miscellaneous consumption items like self-productionand meals bought from outside. Besides food expenditures, there are generallyseveral other categories of expenditures included in the national consumptionaggregates. Table 6 summarizes the other food and nonfood items that South Asiancountries include as part of the construction of the national consumption aggregates.

All countries include self-production in the national consumption aggregate,but the questionnaire design to extract information on home production variesacross countries. Most countries ask separate questions about the value of homeproduction and the value of either market or total consumption for each item. InAfghanistan, Bangladesh, and Maldives, however, households are only asked aboutthe value of total item consumption and are asked to identify whether the primarymode of acquisition was through the market or home production.

With respect to food away from home, Bangladesh, Pakistan, and Sri Lankado not include food or meals bought from outside the household as part of thenational consumption aggregate, but other countries do. In Maldives, the surveyasks the quantity of and expenditure on meals bought outside the household, whileNepal asks how many months in the past year the household purchased food fromoutside and the total estimated amount spent on it. Bhutan asks the number of timesthe members eat out in a month, number of those meals they pay for, and average

8For a general discussion of the issue and policy recommendation on questionnaire design, see alsoBrowning, Crossley, and Weber (2003) and Gibson (2006).


Tabl

e6.

Sum

mar

yof

Oth

erFo

odan

dN

onfo

odIt

ems

Incl

uded

inth

eN

atio

nalC

onsu

mpt

ion

Agg

rega

tes

Afg

hani

stan

Ban

glad

esh

Bhu

tan

Indi

aM

aldi

ves

Nep

alP

akis

tan

SriL

anka

Food

expe

ndit

ures

Isse

lf-p

rodu

ctio

nac

coun

ted

for?

aY

esa

Yes

aY

esb

Yes

Noa

Yes

Yes

Yes

Are

mea

lsou

tsid

eth

eho

useh

old

acco

unte

dfo

r?Y

esN

ocY

esd

Yes

eY

esf

Yes

gN

oN

o

Non

food

expe

ndit

ures

Num

ber

ofit

ems

incl

uded

193

8528

640

162

7447

Num

ber

ofit

ems

excl

uded

1026

028

3531

35A

reco

nsum

erdu

rabl

esac

coun

ted

for?

Yes

Yes

hY

esh

Yes

hN

oY

esN

oY

esh

Hou

sing

Act

ualr

enti

nclu

ded?

Yes

Yes

Yes

Yes

No

Yes

Yes

Yes

Impu

ted

rent

incl

uded

?Y

esY

esi

Yes

iN

oN

oY

esj

Yes

iY

esi

Hea

lth

and

educ

atio

nH

ealt

hex

pend

itur

es?

No

Yes

Yes

Yes

Yes

No

Yes

Yes

Edu

cati

onex

pend

itur

es?

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Oth

ers

Mai

nit

ems

excl

uded

Wed

ding

s,ce

lebr

atio

ns,

dona

tion

s,ta

lism

an,a

ndot

her

mis

cell

aneo

usex

pend

itur

es

Occ

asio

nal

expe

ndit

ures

,in

com

eta

x,in

tere

stch

arge

s,an

din

sura

nce

Agr

icul

tura

lin

put,

tax

and

insu

ranc

e,an

dot

her

Non

eO

ccas

iona

lex

pend

itur

e,du

rabl

es,r

ent,

mor

tgag

e,an

dlo

anbr

oker

age

serv

ices

Occ

asio

nal

expe

nses

kO

ccas

iona

lex

pend

itur

es,

dura

bles

,tax

,an

dan

nual

item

sl

Occ

asio

nal

expe

ndit

ures

,ta

x,in

sura

nce,

and

cont

ribu

tion

sfu

nds,

dura

bles

CO

ICO

P=

clas

sifi

cati

onof

indi

vidu

alco

nsum

ptio

nac

cord

ing

topu

rpos

e.a To

talc

onsu

mpt

ion,

incl

udin

gse

lf-p

rodu

ctio

n,is

repo

rted

inon

equ

esti

onpe

rit

em.I

not

her

coun

trie

s,m

arke

tand

self

-pro

duct

ion

are

repo

rted

sepa

rate

lyfo

rea

chit

em.

bS

elf-

prod

ucti

onis

also

acco

unte

dfo

ram

ong

nonf

ood

item

s(B

lock

10of

the

ques

tion

nair

e).

c Que

stio

nnai

re:S

ecti

on9A

2it

em14

.dQ

uest

ionn

aire

:Var

iabl

esat

the

end

ofB

lock

8.e Q

uest

ionn

aire

:Blo

ck12

,ser

ved

proc

esse

dfo

odan

dpa

ckag

edpr

oces

sed

food

.f C

OIC

OP

:111

1.gS

ecti

on5:

CO

ICO

P13

1.hN

oco

nsum

ptio

nfl

owes

tim

ated

,exp

endi

ture

item

sin

clud

edat

purc

hase

pric

e.i Q

uest

ionn

aire

repo

rts

mar

kete

quiv

alen

tval

ues.

j Que

stio

nnai

rere

port

sm

arke

tequ

ival

entv

alue

s;ou

tlie

rsan

dm

issi

ngim

pute

dus

ing

pred

icti

onfr

omhe

doni

cre

gres

sion

s.kS

ome

ofth

ein

freq

uent

expe

nses

have

been

kept

,mos

tnot

ably

lega

lexp

ense

san

din

sura

nce.

l Ann

uall

icen

sefe

es(e

.g.,

tele

visi

on,v

ideo

cass

ette

reco

rder

,and

dish

ante

nna)

;ann

uall

icen

sefe

efo

rar

ms.

Sou

rce:

Aut

hors

’es

tim

ates

base

don

Sou

thA

sia

Har

mon

ized

Mic

roD

atas

et(a

cces

sed

Sep

tem

ber

15,2

017)

.


price of each meal. On the other hand, India includes the number of meals eachmember of the household consumes per month away from home and the price ofcooked meals purchased during the past 30 days.

Accounting for meals eaten outside of the household can have hugeimplications on poverty rates. A study from Peru showed that including theconsumption of food eaten outside the home raises the extreme poverty rate by18% (Farhan, Genoni, and Vakis 2015). Furthermore, most household surveys lackproper data collection methods for food eaten outside the home. In a recent analysisof household surveys globally, 90% of the surveys asked about food eaten awayfrom home, but few had any additional follow-up questions (Smith, Dupriez, andTroubat 2014). Only 23% of the surveys in the study collected expenditure dataon food eaten away from home, and 17% collected data on meals eaten outsidethe home at the individual level. This becomes an increasingly significant issue ascountries develop and a larger share of food is consumed outside the home.

Nonfood, Nondurable Items

Most countries include nonfood, nondurable items, but the number of itemsincluded varies from country to country. Sri Lanka includes just 47 items, whileBhutan, Nepal, and Pakistan include between 60 and 85 items. On the other hand,Bangladesh and India include over 190 items, and Maldives lists 401 items.

Consumer Durables

When it comes to consumer durable goods, Maldives and Pakistan do notinclude them in the consumption aggregate, but the rest of the countries do. InAfghanistan and Nepal, the monthly value of the consumption of durables isimputed, following recommended practice (Deaton and Zaidi 2002). In Bangladesh,Bhutan, India, and Sri Lanka, however, only the previous year’s expenditures onconsumer durables are included.

Housing Expenditures

All countries in South Asia include actual rent for urban and rural areas,except for India and Maldives, which do not include imputed rent for homeowners.However, Deaton and Zaidi (2002), in their comprehensive study about howto construct consumption aggregates, recommended including imputed rent forhomeowners in their consumption aggregate and only excluding it in extreme cases.The India NSS asks urban homeowners, but not rural homeowners, to estimate rent.Although the NSS collects imputed rent from urban dwellers, this is not included inthe consumption aggregate of urban dwellers.


Health and Education Expenditures

All countries except for Afghanistan and Nepal have an estimate of healthexpenditures included in the national consumption aggregate. All countries includeeducation expenditures in their national aggregate.

IV. Spatial Deflation to Account for Cost-of-Living Differencesand Intertemporal Deflation

Spatial Deflation of National Consumption Aggregatesto Account for Cost-of-Living Differences

Spatial deflation is an important requirement to properly assess the numberof poor by adjusting for cost-of-living differences across geographical areas so thatpoverty is not overstated in low-cost areas. For example, prices are generally higherin urban areas compared to rural areas, meaning that an urban household wouldneed to spend more to maintain the same standard of living as that of a ruralhousehold. Failing to adjust for cost-of-living differences overestimates poverty inrural areas and underestimates it in urban areas. Empirically, many studies haveshown the importance of using regional price indexes. For example, Hill and Syed(2015) show that using urban–rural price differentials reduces the gross domesticproduct estimate of the People’s Republic of China (PRC), and Weinand and vonAuer (2020) show a wide dispersion of prices in different regions of Germany.

However, it could be difficult to collect regional prices for all goods andservices. To circumvent this issue, some research have promoted the use of theEngel curve approach to deflate expenditures across regions and time instead ofusing the national CPI to deflate prices, leading to a more accurate spatial deflator.9

For example, Almas and Johnsen (2012) showed that using regional price indexes todeflate expenditures in the PRC resulted in a modest reduction in national povertyover time and a larger increase in inequality. They also found that using the regionalprice deflator understates the fall in the $1 per day poverty measurement in ruralareas compared to using CPI as a price deflator, but not the $2 per day povertymeasurement. Using the Engel curve to deflate prices in India would overstate thenational poverty rate in that country by about 4 percentage points in both rural andurban areas (Almas, Kjelsrud, and Somanathan 2019). However, the Engel curveapproach is not foolproof; the example of Viet Nam shows that it is not a good

9The Engel curve approach states that expenditure on food as a proportion to total income should fall asincomes rise. That is, if in two similar households the proportion of income spent on food is the same, then their realincomes must be equal. Using this approach, the regional price indexes are estimated using household food shares asthe dependent variable, and demographic controls, income, and regional dummies as regressors. The coefficients ofthe regional dummies give regional price indexes.


proxy of price indexes that are constructed from actual regional prices (Gibson, Le,and Kim 2017; Gibson and Le 2018).

The other issue is whether housing costs should be included in spatialcost-of-living adjustments and how to incorporate the rental value of owner-occupied housing. Almas and Johnsen (2012) use 8% of housing value as a proxyfor rent and adjust for housing debt and subsidies to obtain the true cost of housing.They find that including imputed rent into a spatial deflator reduced rural povertyrates in 1995 compared to using spatial deflators without imputed rent. However,rural poverty rates in 2002 and urban poverty rates in both 1995 and 2002 werevirtually the same regardless of whether the spatial deflator had imputed rentalvalues or not. Gibson, Le, and Kim (2017) used a hedonic price model to estimaterental values and found that including the imputed rental values has a marginaleffect on the spatial price indexes.

International poverty rates in the region in most cases are based on welfareaggregates that are not spatially deflated. While all South Asian countries carryout spatial deflation when calculating their national poverty estimates, the WorldBank currently only spatially deflates the welfare measures in Bhutan and Nepalwhen calculating international extreme poverty rates. Bhutan uses a survey-basedprice index to deflate prices. In this case, the spatial price indexes are derivedusing information from the survey. Similarly, Bangladesh and Nepal use an implicitspatial price index to deflate prices, which is constructed using region-specificpoverty lines to determine the cost-of-living differences (Deaton and Muellbauer1980).10

Table 7 shows the international poverty rates using the nominal and spatiallydeflated national consumption aggregates. Overall, the absence of spatial deflationin calculating extreme poverty rates might have a minor effect on the estimationof country-level international poverty rates, but it can have substantial impactson international poverty measurement across regions and within countries. In allSouth Asian countries except for Nepal, the impact of spatial deflation is nontrivial.Without spatial deflation, urban areas have less poverty and rural areas have morepoverty. For example, spatial deflation causes Nepal’s urban poverty rate to be 4.3percentage points higher and rural poverty rate to be 7.1 percentage points lower,compared with their respective nominal rates. If Bangladesh spatially deflated itsconsumption aggregates, then it would see an urban extreme poverty rate 3.3percentage points higher and a rural extreme poverty rate 4.7 percentage pointslower.

10As an illustrative example, consider a country with two regions. The poverty line in a region can beinterpreted as the cost of achieving a reference utility level in that region. The ratio between the poverty lines intwo regions can be interpreted as a true-cost-of-living index—that is, a spatial price index. Thus, the ratio betweenthe regional poverty lines gives the cost of achieving the reference utility level in region 1 relative to region 2. Thismethod makes strong assumptions including that the unobserved reference utility level underlying the poverty linesis constant across regions.


Table 7. International Extreme Poverty Rate in Urban and Rural Areas, With andWithout Spatial Deflation

International PovertyRate (%)

Urban internationalPoverty Rate (%)

Rural InternationalPoverty Rate (%)

NominalSpatiallyDeflated Nominal

SpatiallyDeflated Nominal

SpatiallyDeflated

Bangladesh 18.5 16.0 6.7 10.0 22.8 18.1Bhutan 2.2 2.2 0.21 0.2 3.4 3.1India 21.6 19.4 13.4 26.4 24.8 16.5Maldives 7.3 7.3 3.9 5.8 10.2 9.2Nepal 20.0 14.9 4.7 9.0 23.4 16.3Pakistana 8.3 7.0 2.8 3.1 10.7 9.0Sri Lanka 1.9 1.7 0.3 0.4 2.3 2.0

aPakistan’s values are from 2010.Source: Authors’ estimates based on South Asia Harmonized Micro Dataset (accessed September 15, 2017).

Therefore, following Gibson, Le, and Kim (2017), it would be beneficial tocollect regional price data for different durable and nondurable goods, and services.The rental cost of housing should also be collected regionally to impute the rentalrate of owner-occupied housing, following Weinand and von Auer (2019), whoshow that housing cost is a major driver of cost-of-living variation. While thehousehold expenditure surveys collect consumption amounts and total expendituresfor each food and nonfood item at the household level, the market prices of the itemsare not generally collected. Thus, to get a better estimate of regional price deflators,statistical agencies need to incorporate the collection of market prices of differentitems in the household surveys.

Intertemporal Deflation

As the survey year may not align with the year the $1.9 internationalpoverty line was estimated (2011), the international poverty line might need to beintertemporally deflated before it is applied to the national consumption aggregatesof some countries. To do so, the international poverty line set in 2011 dollarsis converted to local currency using the 2011 purchasing power parity (PPP)conversion factor. This number is then intertemporally deflated to the year thesurvey was undertaken, usually using the respective CPI of each country.

Table 8 lists the country-level CPI and PPP that are used to intertemporallydeflate the international poverty line to local currency units. For example, tocalculate the international poverty line expressed in Bangladeshi takas for the year2010 so that it can be applied to the 2010 HIES, we do the following. First, the2011 PPP for Bangladesh (24.85) is multiplied by the international poverty line($1.9), and this is then deflated using the intertemporal deflator (0.9). Thus, the


Table 8. Intertemporal Deflators

Country Year 2011 PPP Intertemporal Deflator

Afghanistana 2012 na naBangladesh 2010 24.85 0.90Bhutan 2012 16.96 1.11India 2011–2012 13.78 1.04

Rural 2011–2012 12.91 1.03Urban 2011–2012 15.69 1.04

Maldives 2009 10.68 0.87Nepal 2010 25.76 0.93Pakistan 2011–2012 25.41 1.05Sri Lanka 2012–2013 42.22 1.11

na = not available, PPP = purchasing power parity.aAfghanistan is not included in PovcalNet because it lacks a PPP deflator, andthe government is not comfortable using a regression-based PPP.Note: All countries use the consumer price index except for Bangladesh, whichuses the Basic Need Price Index.Source: Authors’ estimates based on South Asia Harmonized Micro Dataset(accessed September 15, 2017).

daily international poverty line expressed in local currency comes out to be: 1.9 ×24.85 × 0.9 = 42.65 takas.

All countries in South Asia except for Afghanistan and Bangladesh use theCPI to deflate the international poverty line. Bangladesh uses the Basic Need PriceIndex that is constructed based on Bangladesh’s population-weighted upper povertylines to measure changes in the cost of basic needs. This methodological approachwas adopted for Bangladesh because Gimenez and Jolliffe (2014) argued that theweights used to construct the CPI are not representative of the consumption patternof the poor. This view aligns with Recommendation No. 9 of the World Bank’s(2017) report on Monitoring Global Poverty, which specifies that countries shouldconsider using a price index for the poor. It is also consistent with the belief thatprice changes implied by the Bangladesh survey data are closer to changes in thepoverty line than the CPI.

While the CPI shows how the expenditure to purchase a specific bundleof goods consumed by the average consumer changes nationally, there are a fewproblems with using the CPI as a deflator for poverty measurement: (i) unit-valuechanges of some consumption items can vary from the CPI for this item; (ii) theCPI measures how price changes nationally and not by region, so it does not reflectprice changes of specific regions of the country; or more importantly, it may notaccurately identify the poor in regions with high levels of deprivation; and (iii) thegoods in the poor’s basket may not align well with the CPI basket, as the extremelypoor may consume items that are different from that of a typical consumer. This lastpoint is validated by Dupriez (2007) and Deaton and Dupriez (2009, 2011), whodocument that the expenditure patterns of poor households are different from the


pattern observed in the System of National Accounts.11 One limitation of our studyis that we do not have access to satisfactory country-level price deflators that reflectthe consumption bundle of the poor (because most governments do not producesuch data), so we have to rely on the country-level price deflators instead to deflatethe consumption expenditures. However, a recent study by Deaton and Dupriez(2011) investigating how sensitive world poverty numbers are to different deflators(i.e., PPP exchange rates and poverty-weighted PPPs) finds that the worldwideinternational poverty numbers are not very sensitive to the choice of the deflator.12

Deaton and Dupriez (2011) and Castañeda et al. (2016) show that subnationalspatial deflation is not the main factor in explaining cross-country differences inpoverty in South Asia. In fact, whether or not one spatially deflates aggregateshaving only minor effects on national poverty rates in South Asia gives our povertyexercise practical validity.13

This section has shown that there are many sources that can substantiallycontribute to the total error of poverty estimates and influence international povertycomparisons. The comparison of survey design and sampling methodologies showsthat differences in data collection methods across countries are relatively minor.The main source of total error involves differences in how consumption is collectedand aggregated at the household level, particularly the way in which countries treatactual and imputed rent. In addition, the use of poverty lines to deflate intertemporalprices instead of the CPI can also contribute to total error in the internationalpoverty rate estimates in the case of Bangladesh.

V. Assessing Differences in Estimating National Consumption Aggregates

Standardized versus National Consumption Aggregates

Because of the wide variation in the way consumption data are collected byeach country, the World Bank developed a standardized dataset of consumptionaggregates.14 The standardization of the consumption aggregates was constructedby reclassifying expenditure items into the categories adopted by the InternationalComparison Program—which has 110 basic headings, 91 classes, 43 groups, and 13categories—and designing strict data cleaning procedures to ensure consistency in

11Sen (1979) offers a thorough review of the pitfalls of using different price deflators (Laspyeres and PaascheIndexes) to obtain real income for welfare analysis.

12The paper measures the PPP exchange rates and the poverty-weighted PPPs using the Fisher Ideal Index,which is the geometric mean of the Laspeyres and Paasche Indexes, as well as the Tȍrnqvist Index and the weightedcountry-product-dummy index.

13We also acknowledge if we are not measuring national poverty rates and instead focus on analyzingsubnational poverty rates, then a local price deflator would be more appropriate.

14As mentioned previously, we use the term national consumption aggregates to refer to the consumptionaggregates created by the statistical offices of the respective countries, while the standardized consumption aggregatesrefer to the consumption aggregates obtained from the standardized consumption datasets created by the World Bank.


the standardization.15 However, this standardization has limitations (Dupriez 2007).Although consumption aggregates are standardized in the sense that they use acommon data dictionary, the method of data collection and questionnaire designsaffect the standardization directly and cannot be addressed.

Table 9 shows the average per capita consumption using both thestandardized and national consumption aggregates.16 The table shows that, for themost part, there is not a large difference in expenditure per capita between the twoaggregates. One exception is housing expenditure in urban India. Using the nationalconsumption aggregates, imputed rent amounts to $0.7 per day (2011 PPP), butwhen imputed rents for homeowners are included in the standardized consumptionaggregate, the housing expenditure jumps to $1.6 (2011 PPP). The standardizationof the consumption aggregates decreases housing expenditure in Bhutan and SriLanka. Sri Lanka also sees a fall in food and nonfood (nondurable) expendituresdue to the standardization process.

Table 10 shows the budget shares for each category of goods and servicesusing both the standardized and national consumption aggregates. Maldivianhouseholds only spend 37% of consumption expenditures on food, on average,and 53% on nonfood nondurables.17 Households in the remaining countries spendbetween 44% and 59% of their consumption expenditures on food and somewherebetween 18% and 29% on nonfood nondurables. Except for India, all SouthAsian countries spend around 10% of their consumption expenditures on health,education, and durable goods combined, with India spending at a slightly higherrate of 12%. Housing, which includes rent, imputed rent (if included in the nationalconsumption aggregate), and expenditures on utilities range widely from a low of6% in Nepal to a high of 33% in Maldives.

Within India, rural households spend 8 percentage points more on food,reflecting lower levels of well-being in rural India. There is some difference betweenhousing expenditures: while rural Indian households report spending an average of12% of their budget on housing, urban Indian households spend an average of 16%(excluding imputed rent). This implies that much of the difference in poverty ratesbetween rural India and the rest of South Asia is driven by the lack of housingexpenditure (imputed rent) data in the national consumption aggregate for ruralIndia.

15A detailed description of the standardization of the consumption aggregates in South Asia is available inDupriez (2007).

16In Table 9 and subsequent tables, we use the Uniform Reference Period (URP), which collects expendituredata on all goods over a 30-day period. India also reports the Mixed Reference Period, in which food expendituresare collected in 7- and 30-day periods, and other nonfood consumption is collected over a 365-day period (WorldBank 2018). The World Bank currently uses the URP method of collecting expenditure data to measure consumptionexpenditures. Thus, we use the URP method here to determine household-level consumption. However, from 2020onward, the World Bank will use the Modified Mixed Reference Period, where expenditures will be collected over arecall of 7-, 30-, and 365-day periods, depending on the category of the good (World Bank 2018).

17These are plutocratic rather than democratic shares.


Tabl

e9.

Ave

rage

Dai

lype

rC

apit

aE

xpen

ditu

reby

Cat

egor

yof

Goo

dsan

dSe

rvic

es($

)

Ban

glad

esh

Indi

a

All

Rur

alU

rban

Bhu

tan

All

Rur

alU

rban

Mal

dive

sN

epal

Pak

ista

nSr

iLan

ka20

1020

1020

1020

1220

11–2

012

2011

–201

220

11–2

012

2009

2010

2011

–201

220

12–2

013

Nat

iona

lcon

sum

ptio

nag

greg

ates

Food

1.9

1.8

2.4

2.5

1.6

1.5

1.9

1.8

2.0

2.0

3.2

Non

food

nond

urab

les

0.8

0.7

1.1

2.5

1.0

0.8

1.4

3.3

1.5

0.9

2.9

Hea

lth

0.1

0.1

0.2

0.5

0.3

0.2

0.3

0.5

0.1

0.3

Edu

cati

on0.

30.

20.

50.

20.

20.

10.

40.

20.

20.

10.

3D

urab

lego

ods

0.1

0.1

0.1

0.1

0.5

0.4

0.7

0.3

0.2

Hou

sing

0.6

0.4

1.0

1.7

0.4

0.3

0.7

1.6

0.3

0.9

2.0

Tota

l3.

73.

15.

27.

54.

13.

55.

67.

44.

34.

08.

8S

tand

ardi

zed

cons

umpt

ion

aggr

egat

esFo

od1.

81.

62.

22.

51.

51.

51.

71.

72.

01.

82.

7N

onfo

odno

ndur

able

s0.

70.

61.

02.

21.

10.

91.

52.

10.

91.

01.

6H

ealt

h0.

10.

10.

10.

30.

20.

20.

20.

50.

40.

10.

3E

duca

tion

0.2

0.1

0.3

0.1

0.1

0.1

0.2

0.2

0.2

0.2

0.3

Dur

able

good

s0.

10.

00.

10.

40.

10.

10.

20.

90.

50.

10.

3H

ousi

ng0.

50.

40.

91.

30.

70.

31.

61.

80.

30.

81.

1To

tal

3.3

2.9

4.6

6.7

3.7

3.0

5.4

7.1

4.2

4.0

6.0

Not

es:A

llam

ount

sar

eba

sed

on20

11pu

rcha

sing

pow

erpa

rity

.Ind

ia’s

cons

umpt

ion

aggr

egat

esre

port

edar

eba

sed

onth

eU

nifo

rmR

ecal

lPer

iod.

Sou

rce:

Aut

hors

’es

tim

ates

base

don

Sou

thA

sia

Har

mon

ized

Mic

roD

atas

et(a

cces

sed

Sep

tem

ber

15,2

017)

.


Tabl

e10

.B

udge

tSh

ares

byC

ateg

ory

ofG

oods

and

Serv

ices

(%)

Ban

glad

esh

Indi

a

All

Rur

alU

rban

Bhu

tan

All

Rur

alU

rban

Mal

dive

sN

epal

Pak

ista

nSr

iLan

ka20

1020

1020

1020

1220

11–2

012

2009

2010

2011

–201

220

12–2

013

Nat

iona

lcon

sum

ptio

nex

pend

itur

eFo

od58

6153

4551

5446

3757

5245

Non

food

nond

urab

les

1919

2028

2524

2753

2721

25H

ealt

h4

43

45

65

70

32

Edu

cati

on5

48

34

24

35

22

Dur

able

good

s1

11

<1

33

30

40

1H

ousi

ng14

1318

1913

1216

336

2122

Tota

l10

010

010

010

010

010

010

010

010

010

010

0S

tand

ardi

zed

cons

umpt

ion

aggr

egat

eFo

od57

5952

4446

5134

2953

4850

Non

food

nond

urab

les

2020

2030

2828

2828

2023

24H

ealt

h3

33

35

54

79

35

Edu

cati

on6

58

22

23

34

35

Dur

able

good

s2

22

32

22

127

23

Hou

sing

1413

1919

1712

2921

821

18To

tal

100

100

100

100

100

100

100

100

100

100

100

Not

es:I

ndia

’sco

nsum

ptio

nag

greg

ates

repo

rted

are

base

don

the

Uni

form

Rec

allP

erio

d.S

ourc

e:A

utho

rs’

esti

mat

esba

sed

onS

outh

Asi

aH

arm

oniz

edM

icro

Dat

aset

(acc

esse

dS

epte

mbe

r15

,201

7).


Table 11. International Extreme Poverty Rate in South Asia Using theStandardized and National Consumption Aggregates

StandardizedConsumption

Aggregates

NationalConsumption

AggregatesDifference(1) − (2)

Year (1) (2) (3)

Bangladesh 2010 19.44 18.52 0.92Rural 2010 23.86 22.74 1.12Urban 2010 7.01 6.67 0.34

Bhutan 2012 3.49 2.17 1.32India 2011–2012 19.61 21.56 –1.95

Rural 2011–2012 24.20 24.83 –0.63Urban 2011–2012 8.13 13.38 –5.25

Maldives 2009 6.10 7.25 –1.14Nepal 2010 14.51 14.89 –0.38Pakistan 2011–2012 9.24 7.93 1.31Sri Lanka 2012–2013 2.06 1.92 0.14South Asia 18.31 19.60 −1.29

Notes: India’s consumption aggregates reported are based on the Uniform Recall Period.Source: Authors’ estimates based on South Asia Harmonized Micro Dataset (accessed September15, 2017).

Standardization leads to a slight change in food expenditures in all countries.With nonfood nondurables, standardization slightly increases the share of nonfoodnondurables in Bangladesh, Bhutan, India, and Pakistan, but decreases the sharein other countries. Standardization does not affect health care expenditures, andthe proportion of housing expenditure changes little with standardization. However,Indian housing expenditures are much higher than those in the national consumptionaggregate; this is driven mainly by the inclusion of imputed rent among urbanhouseholds. To study this point further, we reproduce Table 8 and Table 9, but onlyfor the poor in each country. The reproduced tables, which are presented in theAppendix as Table A.1 and Table A.2, respectively, show that much of the increasein consumption expenditures among the poor in urban India is driven by imputedrent.

While standardization of consumption aggregates does not affect theinternational poverty rankings for most countries, it has a significant effecton the poverty numbers for India. Table 11 shows the international extremepoverty rates calculated using national consumption aggregates and standardizedconsumption aggregates. In some cases, standardization increases the poverty rate,like in Bangladesh, Bhutan, Pakistan, and Sri Lanka. However, in the cases ofIndia, Maldives, and Nepal, the standardization process decreases the internationalextreme poverty rate.

For the most part, the standardization process preserves the poverty rankings.When we use the national consumption aggregates to estimate the internationalextreme poverty rate, Sri Lanka has the lowest proportion of individuals classified


Figure 1. Sensitivity of the International Extreme Poverty Rate When National andStandardized Consumption Aggregates Are Used to Measure Poverty

BAN = Bangladesh, BHU = Bhutan, IND = India, MLD = Maldives, NEP = Nepal, PAK = Pakistan, PPP =purchasing power parity, SRI = Sri Lanka.Note: The vertical line shows where the international poverty line lies.Source: Authors’ estimates based on South Asia Harmonized Micro Dataset (accessed September 15, 2017).

as extremely poor, followed by Bhutan, Maldives, Pakistan, Nepal, Bangladesh, andIndia. However, when we use the standardized consumption aggregates to do thesame exercise, Bangladesh and India have virtually the same extreme poverty rate,whereas the extreme poverty rate is 3 percentage points higher in India when usingthe national aggregates. The international extreme poverty rate for South Asia as aregion also falls by 1.3 percentage points, when using the standardized consumptionaggregates, from 19.6% to 18.3%.

Figure 1 shows how standardization affects the distribution of per capitaconsumption in each South Asian country. For India, standardization has asignificant impact on the distribution of the consumption aggregates, with thedistribution shifting to the right and away from the $1.9 poverty line. On the otherhand, it has a smaller impact on the distribution of consumption aggregates inBangladesh and Pakistan, shifting them very slightly to the left. This reinforces theobserved decline in India’s poverty rates when using the standardized aggregatesand the small increases in poverty rates for Bangladesh and Pakistan.

Using the standardized national consumption aggregates significantlyreduces the number of poor in India and the region. Table 12 presents the estimatesof the number of international extreme poor in each country using the nationalconsumption aggregates and the standardized consumption aggregates.

Standardization of the consumption aggregates would reduce theinternational extreme poverty rate in South Asia by 1.3 percentage points, oraround 18.5 million people, with India witnessing a reduction of about 22 million.Other countries, like Bangladesh and Pakistan would see a large increase in their


Table 12. Number of Extreme Poor Using Standardized and NationalConsumption Aggregates

Number of Extreme Poor(million)

StandardizedConsumption

Aggregates

NationalConsumption

AggregatesDifference(1) − (2)

Country Year (1) (2) (3)

Afghanistan 2012 na na naBangladesh 2010 28.87 27.49 1.37

Rural 2010 26.12 24.89 1.23Urban 2010 2.74 2.60 0.13

Bhutan 2012 0.02 0.01 0.01India1 2011–2012 217.52 239.10 −21.58

Rural 2011–2012 191.75 196.70 −4.95Urban 2011–2012 25.76 42.40 −16.64

Maldives 2009 0.02 0.02 <0.00Nepal 2010 4.09 4.20 −0.11Pakistan 2010–2011 11.96 10.26 1.69Sri Lanka 2012–2013 0.43 0.38 0.05Total 262.91 281.47 −18.55

na = not available.Sources: Population numbers obtained from World Bank. PovcalNet. http://iresearch.worldbank.org/PovcalNet/povOnDemand.aspx (accessed September 25, 2017); Authors’ estimates based on SouthAsia Harmonized Micro Dataset (accessed September 15, 2017).

respective poverty numbers if consumption aggregates were standardized. Pakistanwould see an increase in the number of extreme poor from 10.3 million to 12million, and Bangladesh would see its number of extreme poor rise from 27.5million to 28.9 million, mainly driven by the rural poor. As mentioned, Indiawould experience a decrease in the number of extreme poor of about 22 millionnationwide. Most of the fall would result from the decrease in the number of urbanpoor in India, due to the inclusion of imputed housing rent for homeowners inthe standardized welfare aggregate. However, as the length of recall can affectconsumption expenditures, harmonizing the length of recall in different SouthAsian countries could have some effect on these poverty numbers.

VI. Consistency of Monetary International Poverty Rates with NonmonetaryWelfare Dimensions

This section considers whether the international extreme poverty raterankings are consistent with nonmonetary dimensions of welfare. Thus far, wehave examined monetary metrics of individual well-being, including the nationaland standardized consumption aggregates. This section assesses how the rankingof monetary welfare relates to other nonmonetary dimensions of well-being. Sen(1979) introduced the concept of capabilities, which states that a person has a set

http://iresearch.worldbank.org/PovcalNet/povOnDemand.aspx


of objectives to choose from to lead a meaningful, valued life. Many reports andpapers have used Sen’s capability approach to highlight the importance of reportingdifferent indicators other than income to measure the overall well-being of a person.For example, Bourguignon and Chakravarty (2019) stressed the importance ofmeasuring nonmonetary indicators of well-being because even though a personmay have enough income to afford higher consumption, problems arise when themarkets for certain nonmonetary welfare indicators either do not exist or are notperfect. Thus, different nonmonetary factors must be assessed besides income(e.g., access to clean water and educational attainment) to get a broader pictureof well-being. The World Bank recently released Poverty and Shared Prosperity2018: Piecing Together the Poverty Puzzle, which closely examines nonmonetaryindicators to assess the poverty status of a person. The report also computesthe Multidimensional Poverty Index of Alkire and Foster (2011), an index thatamalgamates different monetary and nonmonetary indicators of well-being into onemeasure of poverty.

A country with high levels of (monetary) international poverty may performbadly in other nonmonetary welfare dimensions. Table 13 lists some indicatorsof nonmonetary well-being for the subpopulation of poor people identified usingthe national consumption aggregates. The nonmonetary indicators in most casesare broadly consistent with the monetary poverty rates, although Bhutan andPakistan are outliers with respect to literacy. Sri Lanka has the lowest proportionof household members that are illiterate, while Pakistan has the highest andBangladesh the second highest. In South Asia, Pakistan scores the worst whenit comes to education among the extreme poor. Although Bhutan has thesecond-lowest extreme poverty rate in the region, its extreme poor are worse off inmany categories when they are compared with India’s extreme poor. For example,only 19% of the heads of households that are extremely poor in Bhutan are literate,compared to 50% in India and 29% in Bangladesh.

Bangladesh scores much worse in many of the nonmonetary dimensionsthan India. India has the highest proportion of extreme poor in the region, whileBangladesh has the second-highest proportion. But the Bangladeshi extreme poorhave less human capital and lower levels of physical assets compared to theirIndian counterparts. Bangladeshi households in extreme poverty have lower levelsof educational attainment and lower ownership rates of household assets, eventhough the extreme poor in both Bangladesh and India have the same proportionof households engaged in agriculture.

Using the standardized consumption aggregates only partly mitigates someof the disparities observed in monetary and nonmonetary indicators. Table 14shows the different nonmonetary indicators of well-being of the extreme poor whenidentified using the standardized consumption aggregates rather than the nationalconsumption aggregates. Bangladesh still scores much worse than India in manyof the dimensions. For example, Bangladeshi extreme poor have lower levels of


Tabl

e13

.Su

mm

ary

Stat

isti

csof

Indi

cato

rsof

Non

mon

etar

yW

ell-

Bei

ngfo

rth

eE

xtre

me

Poo

rId

enti

fied

Usi

ngth

eN

atio

nal

Con

sum

ptio

nA

ggre

gate

s

Var

iabl

eB

angl

ades

hB

huta

nIn

dia

Mal

dive

sN

epal

Pak

ista

nSr

iLan

ka

Hou

seho

ldco

mpo

siti

onH

ouse

hold

size

5.73

8.40

6.43

7.53

7.16

10.0

15.

60D

epen

denc

yra

tio

0.97

0.73

0.78

0.67

1.16

1.42

0.70

Edu

cati

onat

tain

men

tin

hous

ehol

dH

ouse

hold

head

know

sho

wto

read

and

wri

te0.

290.

190.

501.

000.

370.

28P

ropo

rtio

nof

hous

ehol

dm

embe

rsw

hoar

eil

lite

rate

0.70

0.63

0.44

0.38

0.63

0.80

0.21

Hou

seho

ldm

embe

rsw

ith

prim

ary

scho

olin

g0.

180.

330.

350.

180.

320.

180.

37R

atio

offe

mal

ech

ildr

enin

scho

olto

allc

hild

ren

insc

hool

0.54

0.41

0.48

0.53

0.35

0.47

Em

ploy

men

tH

ouse

hold

head

enga

ged

inag

ricu

ltur

e0.

560.

550.

370.

490.

53H

ouse

hold

head

self

-em

ploy

ed0.

350.

240.

480.

390.

800.

390.

32H

ouse

hold

asse

tsan

dfa

cili

ties

Hou

seho

ldha

sa

land

line

phon

e0.

010.

010.

120.

020.

010.

09H

ouse

hold

has

elec

tric

ity

0.25

0.58

0.57

0.43

0.74

0.64

Hou

seho

ldha

sa

radi

o0.

050.

360.

160.

690.

410.

020.

45H

ouse

hold

has

ate

levi

sion

0.10

0.12

0.29

0.97

0.16

0.19

0.49

Hou

seho

ldha

sa

fan

0.15

0.06

0.45

1.00

0.14

0.71

0.15

Hou

seho

ldha

sa

sew

ing

mac

hine

0.02

0.01

0.08

0.54

0.05

0.25

0.10

Hou

seho

ldha

sa

bicy

cle

0.15

0.02

0.66

0.34

0.28

0.26

0.36

Dri

nkin

gw

ater

from

hygi

enic

sour

ce0.

960.

960.

730.

870.

84A

vail

abil

ity

ofpr

oper

sani

tati

onfa

cili

ties

0.3

0.61

0.1

0.28

0.74

Not

e:A

llre

port

edsu

mm

ary

stat

isti

csar

epo

pula

tion

wei

ghte

d.S

ourc

e:A

utho

rs’

esti

mat

esba

sed

onS

outh

Asi

aH

arm

oniz

edM

icro

Dat

aset

(acc

esse

dS

epte

mbe

r15

,201

7).


Tabl

e14

.Su

mm

ary

Stat

isti

csof

Indi

cato

rsof

Non

mon

etar

yW

ell-

Bei

ngfo

rth

eE

xtre

me

Poo

rId

enti

fied

Usi

ngth

eSt

anda

rdiz

edC

onsu

mpt

ion

Agg

rega

tes

Var

iabl

eB

angl

ades

hB

huta

nIn

dia

Mal

dive

sN

epal

Pak

ista

nSr

iLan

ka

Hou

seho

ldco

mpo

siti

onH

ouse

hold

size

5.68

8.18

5.60

7.58

7.20

9.68

5.68

Dep

ende

ncy

rati

o0.

470.

400.

950.

410.

530.

570.

41E

duca

tion

atta

inm

enti

nho

useh

old

Hou

seho

ldhe

adkn

ows

how

tore

adan

dw

rite

0.27

0.21

0.50

1.00

0.34

0.31

Pro

port

ion

ofho

useh

old

mem

bers

who

are

illi

tera

te0.

690.

620.

450.

430.

620.

760.

20H

ouse

hold

mem

bers

wit

hpr

imar

ysc

hool

ing

0.18

0.33

0.14

0.17

0.33

0.20

0.35

Rat

ioof

fem

ale

chil

dren

insc

hool

toal

lchi

ldre

nin

scho

ol0.

520.

490.

510.

530.

370.

49E

mpl

oym

ent

Hou

seho

ldhe

aden

gage

din

agri

cult

ure

0.54

0.54

0.59

0.41

0.46

Hou

seho

ldhe

adse

lf-e

mpl

oyed

0.37

0.14

0.42

0.43

0.83

0.37

0.31

Hou

seho

ldas

sets

and

faci

liti

esH

ouse

hold

has

ala

ndli

neph

one

0.01

0.00

60.

070.

020.

020.

13H

ouse

hold

has

elec

tric

ity

0.29

0.65

0.55

0.38

0.82

0.74

Hou

seho

ldha

sa

radi

o0.

050.

350.

140.

650.

430.

020.

48H

ouse

hold

has

ate

levi

sion

0.12

0.22

0.23

0.93

0.13

0.29

0.60

Hou

seho

ldha

sa

fan

0.19

0.06

0.40

1.00

0.12

0.78

0.22

Hou

seho

ldha

sa

sew

ing

mac

hine

0.03

0.01

0.05

0.55

0.04

0.25

0.17

Hou

seho

ldha

sa

bicy

cle

0.16

0.04

0.59

0.30

0.28

0.23

0.35

Dri

nkin

gw

ater

from

hygi

enic

sour

ce0.

960.

980.

750.

860.

87A

vail

abil

ity

ofpr

oper

sani

tati

onfa

cili

ties

0.33

0.61

0.08

0.34

0.79

Not

e:A

llre

port

edsu

mm

ary

stat

isti

csar

epo

pula

tion

wei

ghte

d.S

ourc

e:A

utho

rs’

esti

mat

esba

sed

onS

outh

Asi

aH

arm

oniz

edM

icro

Dat

aset

(acc

esse

dS

epte

mbe

r15

,201

7).


educational attainment and physical assets compared to their Indian counterparts,even though their dependency ratios and the proportion of households engaged inagriculture are similar. This is consistent with the common view that Bangladesh ispoorer than India.

While the international extreme poverty rate shows that India has amuch higher poverty rate than Bangladesh, further analysis of the nonmonetarydimensions of well-being shows that the depth of deprivation is higher inBangladesh than in India, strongly suggesting that the extremely poor in Bangladeshare worse off than those in India. Monetary measures of poverty show that Indiaand Bangladesh have similar rates of poverty when imputed rent is included forurban households, and that India would have a lower extreme poverty rate if imputedhousing rent were included for rural households.

VII. Conclusions

The World Bank has set an ambitious target of eliminating internationalextreme poverty by 2030 and has taken steps to obtain proper estimates of extremepoverty. One of the key challenges in the measurement of poverty is how to identifythe poor. The World Bank uses an international poverty line to identify the extremepoor, which is currently set at $1.9 per person per day in 2011 PPP dollars.Measuring the consumption of goods and services by households is challengingand very sensitive to survey design and data collection methods. Thus, differencesin the way that household surveys are designed and collected across countries canhave significant effects on poverty measurement, which are amplified when doinginternational poverty comparisons.

In this paper, we examine how the construction of consumption aggregatesat the country level in South Asia adds to the total error that is implicitly part ofany international poverty rate estimation. Indeed, when the World Bank recentlyconvened a group of experts to identify how to improve the measurement ofinternational extreme poverty, one of their main recommendations was to provide amargin of error with every poverty estimate. The rationale for this recommendationis that nonsampling errors, or total error, could create a biased estimate ofinternational extreme poverty. Although the region includes only eight countries,it has both the second-highest number and proportion of the world’s extreme poor.Additionally, all South Asian countries use the national consumption aggregate asthe basis to identify the poor. This makes the analysis both tractable and significantin the global context.

For each country, we document in detail the differences in survey designand sampling, highlight the use of spatial deflation and intertemporal deflation, andassess the sensitivity of the international poverty rate to the methodology used toconstruct the consumption aggregates. We find that there is a significant variation inthe methods used to measure poverty across countries in South Asia. Each country


collects a different list of food items from households. For instance, Sri Lankacollects just 47 nonfood nondurable items, while Bangladesh and India include over190 items, and Maldives includes 401 items. Maldives, Pakistan, and Sri Lanka donot add durable goods to the consumption aggregate, but the rest of the countriesin the region do. Afghanistan and Nepal do not include health care expenditures,and India and Maldives do not include imputed rent of owner-occupied housing.Moreover, the methodology to collect consumption data also varies from countryto country. For instance, Maldives does not include home production of goods inthe national consumption aggregate, and Bangladesh, Pakistan, and Sri Lanka donot include meals bought outside the home as part of the national consumptionaggregate.

Still, by far, the most important source of differences in the construction ofconsumption aggregates is the treatment of imputed rents. Except for Maldives, allthe other countries in South Asia collect imputed rent of owner-occupied housingas part of the consumption aggregate. While India does not include imputed rentas part of its national consumption aggregate, it does collect this information forurban dwellers. We find that the absence of imputed rent of owner-occupied housingfrom the national consumption aggregate of India is the main factor that makes itsproportion of international extreme poor the highest in South Asia.

Using the standardized consumption aggregate reduces the number ofextreme poor in South Asia and makes India appear equally poor as Bangladesh.While India has the largest proportion of international extreme poor when thenational consumption aggregates are used to measure poverty, Bangladesh becomeshome to the largest proportion of international extreme poor in South Asia whenstandardized datasets are used. This is mainly being driven by housing expenditures.The standardized dataset includes imputed rent of owner-occupied housing in urbanregions of India. However, some of the results we obtain are affected by thedifferences in recall rates in different countries, and harmonizing them would helpto obtain a more accurate poverty number.

The choice of CPI also plays an important role in determining the number ofextreme poor. Each country deflates the international poverty line to align it withthe year the survey was undertaken. Except for Afghanistan and Bangladesh, allother countries use the CPI to deflate prices. The evidence from Bangladesh showsthat if the CPI were used to measure international extreme poverty (instead of theofficial Basic Need Price Index), then the proportion of the extreme poor would beoverstated.

Spatial deflation has minor impacts on the international poverty ratesmeasured at the country level, but it significantly affects urban and rural povertyrates. Currently, among countries in the region, only Bhutan and Nepal spatiallydeflate their national consumption aggregates when calculating the internationalpoverty rate. The estimation of international poverty rates is not sensitive to spatial


deflation, but this could be an important issue to study subnational internationalpoverty rates.

Nonmonetary measures of poverty are more consistent with the standardizedthan the national consumption aggregates. To understand the standard of living ofthe international extreme poor of each country, we also look at other nonmonetarywelfare indicators. We find that, in some countries, the extreme poor fare better thanin other countries. For example, the Pakistani extreme poor have the lowest levelsof educational attainment, while Sri Lankans have the highest. We also find thatIndia’s extreme poor have better indicators of nonmonetary well-being comparedto the extreme poor of Bangladesh. This matches well with the notion that Indiaoverall is less deprived than Bangladesh.

Our analysis indicates that the ex post harmonization of welfare aggregateswould make international poverty comparisons more accurate. While household-level data provide valuable information at the country level, their idiosyncrasies mayintroduce noise into cross-country comparisons and, importantly, contribute to thetotal error of poverty estimates. Countries collect different numbers of consumptionitems, use different lengths of time for recall, include different categories of goodsin the consumption aggregate, and use CPI to deflate prices, which all affectpoverty estimates. This paper shows how differences in the methodologies of datacollection and calculation of poverty rates across South Asia can affect countries’international poverty rates. These rates change significantly when standardizedwelfare aggregates are used. India and Maldives do not include imputed rent inthe consumption aggregate, but other countries do, and adjusting for this reducesthe number of extreme poor in India substantially. Standardizing the consumptionaggregates, which includes adding imputed rent from urban Indian households,reduces the number of extreme poor in South Asia by almost 18.5 million, or byabout 1.3 percentage points, with India leading the way with a decline of almost21.6 million.

Our study shows that including imputed rent in the consumption aggregatewould substantially improve its measurement and help researchers in getting a moreaccurate assessment of the number of poor. From the perspective of the cost ofcollecting this data, it should not be very difficult or expensive to add one morequestion about imputed rent in the consumption surveys. For example, in the caseof India, the NSS already collects imputed rent data from urban areas, and so, usingthe same questionnaire for rural areas would allow for a harmonized collectionof imputed rent data from across the country. Poverty experts can then includeimputed rent when calculating per capita consumption aggregates to get a moreaccurate measurement of poverty. This endeavor is worthwhile given the largedecline in poverty that we see when imputed rents are included in the consumptionaggregate. However, we also note that poverty experts should study whether anychanges in housing supply can affect imputed rent and whether that could impactthe changes in poverty rates. Thus, in the coming years, it will be important


to explore the effects of adopting a consistent approach to the construction ofwelfare aggregates, particularly with respect to the treatment of housing rent, whenestimating international poverty rates. This can lead to a more accurate assessmentof the world’s progress toward eradicating extreme poverty.

References

Alkire, Sabina, and James Foster. 2011. “Counting and Multidimensional Poverty Measurement.”Journal of Public Economics 95 (7–8): 476–87.

Almas, Ingvild, and Ashild Johnsen. 2012. “The Cost of Living in China: Implications forInequality and Poverty.” Norwegian School of Economics Working Paper.

Almas, Ingvild, Anders Kjelsrud, and Rohini Somanathan. 2019. “A Behavior-Based Approachto the Estimation of Poverty in India.” The Scandinavian Journal of Economics 121 (1):182–224.

Beegle, Kathleen, Joachim De Weerdt, Jed Friedman, and John Gibson. 2012. “Methodsof Household Consumption Measurement through Surveys: Experimental Results fromTanzania.” Journal of Development Economics 98 (1): 3–18.

Bourguignon, François, and Satya R. Chakravarty. 2019. “The Measurement of MultidimensionalPoverty.” In Poverty, Social Exclusion, and Stochastic Dominance, edited by SatyaChakravarty, 83–107. Singapore: Springer.

Browning, Martin, Thomas F. Crossley, and Guglielmo Weber. 2003. “Asking ConsumptionQuestions in General Purpose Surveys.” Economic Journal 113: F540–F567.

Castañeda, Andrés, Dung Doan, David Newhouse, Minh Cong Nguyen, Hiroki Uematsu, andJoão Pedro Azevedo. 2016. Who Are the Poor in the Developing World? Washington, DC:World Bank.

Centre for Urban Studies, National Institute of Population Research and Training, and MeasureEvaluation. 2006. “Slums of Urban Bangladesh: Mapping and Census, 2005.”

Deaton, Angus. 2005. “Measuring Poverty in a Growing World (or Measuring Growth in a PoorWorld).” Review of Economics and Statistics 87 (1): 1–19.

Deaton, Angus, and Olivier Dupriez. 2009. “Global Poverty and Global Price Indexes.”https://www.rrojasdatabank.info/deatonpppnov2009.pdf.

______. 2011. “Purchasing Power Parity Exchange Rates for the Global Poor.” AmericanEconomic Journal of Applied Economics 3: 137–66.

Deaton, Angus, and John Muellbauer. 1980. “An Almost Ideal Demand System.” AmericanEconomic Review 70 (3): 312–26.

Deaton, Angus, and Salman Zaidi. 2002. Guidelines for Constructing Consumption Aggregatesfor Welfare Analysis. Washington, DC: World Bank.

Dupriez, Olivier. 2007. “Building a Household Consumption Database for the Calculation ofPoverty PPPs.” World Bank. Mimeo.

Farhan, Gabriela, María Eugenia Genoni, and Renos Vakis. 2015. “You Are What (and Where)You Eat: Capturing Food Away from Home in Welfare Measures.” World Bank Policy PaperNo. WPS7257.

Ferreira, Francisco H.G., Shaohua Chen, Andrew Dabalen, Yuri Dikhanov, Nada Hamadeh,Dean Jolliffe, Ambar Narayan, Espen Beer Prydz, Ana Revenga, Prem Sangraula, UmarSerajuddin, and Nobuo Yoshida. 2016. “A Global Count of the Extreme Poor in 2012 DataIssues, Methodology, and Initial Results.” The Journal of Economic Inequality 14 (2): 141–72.

https://www.rrojasdatabank.info/deatonpppnov2009.pdf


Gibson, John. 2006. “Statistical Tools and Estimation Methods for Poverty Measures Basedon Cross-Sectional Household Surveys.” In Handbook on Poverty Statistics: Concepts,Methods and Policy Use, United Nations Statistics Division, 128–203. New York.

Gibson, John, Jikun Huang, and Scott Rozelle. 2001. “Why Is Income Inequality So Low inChina Compared to Other Countries? The Effect of Household Survey Methods.” EconomicLetters 71 (3): 329–33.

Gibson, John, and Trinh Le. 2018. “Improved Modelling of Spatial Cost of Living Differences inDeveloping Countries: A Comparison of Expert Knowledge and Traditional Price Surveys.”Working Papers in Economics No. 18/08, University of Waikato.

Gibson, John, Trinh Le, and Bonggeun Kim. 2017. “Prices, Engel Curves, and Time-SpaceDeflation: Impacts on Poverty and Inequality in Vietnam.” The World Bank EconomicReview 31 (2): 504–30.

Gimenez, Lea, and Dean Jolliffe. 2014. “Inflation for the Poor in Bangladesh: A Comparison ofCPI and Household Survey Data.” Bangladesh Development Studies 37 (1/2): 57–81.

Government of Afghanistan, Central Statistics Organization. 2017. “Settled Population by CivilDivisions 2012–2013.” Kabul.

Government of Bangladesh, Bureau of Statistics. 2012. “Census of Slum Areas and FloatingPopulation.” http://203.112.218.65:8008/PageWebMenuContent.aspx?MenuKey=423.

Hill, Robert J., and Iqbal A. Syed. 2015. “Improving International Comparisons of Prices at BasicHeading Level: An Application to the Asia-Pacific Region.” Review of Income and Wealth61 (3): 515–39.

Jolliffe, Dean. 2001. “Measuring Absolute and Relative Poverty: The Sensitivity of EstimatedHousehold Consumption to Survey Design.” Journal of Economic and Social Measurement27: 1–23.

Jolliffe, Dean, Peter Lanjouw, Shaohua Chen, Aart Kraay, Christian Meyer, Mario Negre, EspenPrydz, Renos Vakis, and Kyla Wethli. 2014. A Measured Approach to Ending Poverty andBoosting Shared Prosperity: Concepts, Data, and the Twin Goals. Policy Research Report.Washington, DC: World Bank.

Lanjouw, Jean, and Peter Lanjouw. 2001. “How to Compare Apples and Oranges: PovertyMeasurement Based on Different Definitions of Consumption.” Review of Income andWealth 47: 25–42.

Ravallion, Martin, Shaohua Chen, and Prem Sangraula. 2009. Dollar a Day Revisited.Washington, DC: World Bank.

Sen, Amartya. 1979. “Equality of What.” The Tanner Lecture on Human Values. http://www.ophi.org.uk/wp-content/uploads/Sen-1979_Equality-of-What.pdf.

Smith, Lisa, Olivier Dupriez, and Nathalie Troubat. 2014. Assessment of the Reliabilityand Relevance of the Food Data Collected in National Household Consumption andExpenditure Surveys. Rome: Food and Agriculture Organization of the United Nations andWorld Bank.

UN-Habitat. 2013. “Planning and Design for Sustainable Urban Mobility: Global Reporton Human Settlements.” https://unhabitat.org/sites/default/files/download-manager-files/Planning%20and%20Design%20for%20Sustainable%20Urban%20Mobility.pdf.

Weinand, Sebastian, and Ludwig von Auer. 2020. “Anatomy of Regional Price Differentials:Evidence from Micro-Price Data.” Spatial Economic Analysis 15 (1): 1–28.

World Bank. 2017. Monitoring Global Poverty: Report on the Global Commission on Poverty.Washington, DC.

______. 2018. Poverty and Shared Prosperity 2018: Piecing Together the Poverty Puzzle.Washington, DC.

http://203.112.218.65:8008/PageWebMenuContent.aspx?MenuKey=423

http://www.ophi.org.uk/wp-content/uploads/Sen-1979_Equality-of-What.pdf

https://unhabitat.org/sites/default/files/download-manager-files/Planning%20and%20Design%20for%20Sustainable%20Urban%20Mobility.pdf


App

endi

x

Tabl

eA

.1.

Ave

rage

Exp

endi

ture

per

Day

byC

ateg

ory

ofG

oods

and

Serv

ices

for

the

Poo

r($

)

Ban

glad

esh

Indi

a

All

Rur

alU

rban

Bhu

tan

All

Rur

alU

rban

Mal

dive

sN

epal

Pak

ista

nSr

iLan

ka20

1020

1020

1020

1220

11–2

012

2011

–201

220

11–2

012

2009

2010

2011

–201

220

12–2

013

Nat

iona

lcon

sum

ptio

nag

greg

ates

Food

1.04

1.04

1.03

0.90

0.91

0.92

0.87

0.75

1.05

0.96

0.89

Non

food

nond

urab

les

0.25

0.25

0.26

0.35

0.30

0.29

0.30

0.57

0.31

0.32

0.19

Hea

lth

0.05

0.05

0.04

0.03

0.06

0.06

0.06

0.03

0.05

0.03

Edu

cati

on0.

040.

040.

050.

070.

030.

020.

040.

030.

030.

020.

01D

urab

lego

ods

0.01

0.01

0.01

0.00

0.02

0.02

0.02

0.03

0.01

Hou

sing

0.18

0.18

0.22

0.18

0.21

0.20

0.23

0.17

0.05

0.27

0.32

Tota

l1.

561.

561.

591.

531.

521.

521.

521.

561.

461.

631.

44S

tand

ardi

zed

cons

umpt

ion

aggr

egat

esFo

od1.

001.

000.

980.

830.

860.

870.

770.

730.

970.

921.

08N

onfo

odno

ndur

able

s0.

290.

290.

310.

380.

350.

350.

310.

310.

270.

350.

30H

ealt

h0.

040.

040.

040.

010.

060.

060.

050.

060.

090.

060.

05E

duca

tion

0.04

0.03

0.04

0.05

0.02

0.02

0.03

0.06

0.03

0.03

0.07

Dur

able

good

s0.

010.

010.

010.

010.

010.

010.

020.

140.

040.

010.

02H

ousi

ng0.

180.

180.

220.

250.

230.

200.

390.

250.

110.

270.

18To

tal

1.56

1.56

1.59

1.54

1.52

1.52

1.57

1.54

1.52

1.63

1.69

Not

es:A

llam

ount

sar

eba

sed

on20

11pu

rcha

sing

pow

erpa

rity

.Ind

ia’s

cons

umpt

ion

aggr

egat

esre

port

edar

eba

sed

onth

eU

nifo

rmR

ecal

lPer

iod.

Sou

rce:

Aut

hors

’es

tim

ates

base

don

Sou

thA

sia

Har

mon

ized

Mic

roD

atas

et(a

cces

sed

Sep

tem

ber

15,2

017)

.


Tabl

eA

.2.

Bud

get

Shar

esby

Cat

egor

yfo

rth

eP

oor

(%)

Ban

glad

esh

Indi

a

All

Rur

alU

rban

Bhu

tan

All

Rur

alU

rban

Mal

dive

sN

epal

Pak

ista

nSr

iLan

ka20

1020

1020

1020

1220

11–2

012

2011

–201

220

11–2

012

2009

2010

2011

–201

220

12–2

013

Nat

iona

lcon

sum

ptio

nex

pend

itur

eFo

od67

6765

6060

6058

5671

5961

Non

food

nond

urab

les

1515

1622

1919

1939

2020

12H

ealt

h3

33

24

44

20

32

Edu

cati

on2

23

42

12

24

10

Dur

able

good

s1

11

01

11

02

0H

ousi

ng12

1214

1215

1416

133

1723

Tota

l10

010

010

010

010

010

010

010

010

010

010

0S

tand

ardi

zed

cons

umpt

ion

aggr

egat

eFo

od64

6462

5555

5645

5064

5666

Non

food

nond

urab

les

1818

1923

2424

2318

1722

14H

ealt

h3

32

14

43

46

35

Edu

cati

on2

22

31

12

42

24

Dur

able

good

s1

11

11

11

83

11

Hou

sing

1212

1417

1614

2517

817

17To

tal

100

100

100

100

100

100

100

100

100

100

100

Not

e:In

dia’

sco

nsum

ptio

nag

greg

ates

repo

rted

are

base

don

the

Uni

form

Rec

allP

erio

d.S

ourc

e:A

utho

rs’

esti

mat

esba

sed

onS

outh

Asi

aH

arm

oniz

edM

icro

Dat

aset

(acc

esse

dS

epte

mbe

r15

,201

7).

International Comparisons of Poverty in South Asia

Documents

Transcript of International Comparisons of Poverty in South Asia